This project is an AI-powered healthcare assistant that leverages natural language processing and retrieval-based question answering to respond to users' health-related queries with factual information from reliable sources. Built using LangChain, ChromaDB, and OpenAI's language models, this assistant can interpret medical questions and provide informative, relevant answers.
To build a conversational AI assistant capable of:
- Understanding user queries related to symptoms, illnesses, treatments, etc.
- Fetching accurate answers from a curated set of medical documents.
- Providing context-aware and medically reliable responses in real-time.
- Source: NHS Inform
- Documents: 11 medical documents manually collected and used to build a knowledge base.
- Storage: Embedded and indexed using ChromaDB, a vector store for semantic search.
- LangChain – For building LLM-powered chains.
- OpenAI GPT – For language understanding and generation.
- ChromaDB – For storing document embeddings and retrieving relevant chunks.
- Python – Core development language.
- Jupyter Notebook – Interactive development and testing.
-
Data Ingestion:
- 11 documents from NHS Inform were scraped and prepared.
- Split into smaller chunks using LangChain's text splitter.
-
Embedding:
- Text chunks were embedded using OpenAI Embeddings.
- Stored in ChromaDB for fast semantic retrieval.
-
Query Handling:
- User enters a query in natural language.
- Relevant documents are retrieved from ChromaDB.
- GPT model generates a response based on the retrieved context.
-
Response Display:
- The assistant returns an answer grounded in the documents.